Renal Replacement Therapy Demand Study,
Northern Territory,
2001 to 2022
Jiqiong You
Paul D Lawton
Yuejen Zhao
Susan Poppe
Nicole Cameron
Steven Guthridge
March 2015
Acknowledgements
The authors are grateful to the many people, who have assisted in the production of this report, including:
- Professor Stephen McDonald for his continuing support and for providing survival data from the Australia and New Zealand Dialysis and Transplant Registry;
- Staff from acute care information services and data warehouse for providing hospital data
- Jenny Cleary and Ian Pollock who sponsored this project and provided management support.
© Department of Health, Northern Territory 2015
This publication is copyright. The information in this report may be freely copied and distributed for non-profit purposes such as study, research, health service management and public information subject to the inclusion of an acknowledgement of the source. Reproduction for other purposes requires the written permission of the Chief Executive of the Department of Health, Northern Territory.
Suggested citation
You JQ, Lawton P, Zhao Y, Poppe S, Cameron N, Guthridge S. Renal Replacement Therapy Demand Study, Northern Territory, 2001 to 2022, Department of Health, Darwin, 2015
ISBN 978-0-9757203-3-2
An electronic version is available at:
http://www.health.nt.gov.au/Health_Gains/Publications/index.aspx
General enquiries about this publication should be directed to:
Director, Health Gains Planning Branch
Department of Health
PO Box 40596, Casuarina, NT 0811
Phone: (08) 8985 8074
Email:
Table of contents
Summary v
Introduction 1
Methods 2
Data sources 2
Statistical analysis 2
Descriptive analysis 2
Linear regression model 3
Time-series model 4
Markov chain model 4
Scenario modelling 4
Results 6
Descriptive analysis 6
Models and projections 12
Renal projection overview 12
Time-series model 13
Linear regression model 13
Markov chain model 14
Scenario analysis 17
Scenario 1 – changing the transplant rate 17
Scenario 2 – changing the proportion of the self-care dialysis 17
Scenario 3 – changing the incidence rate per year 17
Scenario 4 – changing the dialysis death rate 17
Scenario 5 – changing the frequency of HD treatments 17
Discussion 21
Appendix: Supplementary tables 26
Abbreviation & glossary 30
References 32
List of tables 33
List of figures 34
Selected Health Gains Planning publications 35
Summary
Across Australia, demand for renal replacement therapy (RRT) services has been growing at a significant rate over the last decade. Health service funders are faced with increasing service delivery costs and investment requirements. In the five-year period from 2007 to 2011, the number of dialysis patients increased by 27% in the Northern Territory (NT). Same day haemodialysis (HD) now comprises close to 50% of total NT public hospital admissions and in recent years the number of NT patients with end-state kidney disease (ESKD) using palliative care has doubled.
This report provides an overview of the projected demand for RRT in the NT in the next ten years. The projections require consideration of four parameters: the number of patients developing ESKD, the proportion of those with ESKD who progress to RRT, the length of time that patients continue using RRT (survival), and the type of RRT. Three separate methods were used for the projections - linear regression, an autoregressive integrated moving average time-series model, and a static Markov chain model.
The number of patients receiving RRT was available from NT hospital data, and around 75% of these patients had been notified to the Australian and New Zealand Dialysis and Transplant Registry (ANZDATA). The ANZDATA registry is used for compiling information about the incidence, prevalence and quality of RRT in Australia and New Zealand. The NT survival rate for dialysis patients has improved substantially in the last ten years. The unadjusted NT median survival improved from 4.5 years in 1995-99 to 6.0 years in 2005-09, which exceeded the recent national median survival (5.0 years). After adjustment for differences in age distribution between NT and Australian populations, the age-adjusted median survival time for the NT of 5.3 years still compared favourably with the national median survival time.
From 2001 to 2012, the number of HD treatments in the NT increased, on average, by around 3,200 per year. The projections for the number of facility-based HD treatments estimate a further increase of between 41% and 70% from 2013 to 2022. The projected average annual increase of HD treatments through this period ranged from 2,700 using the Markov chain model, through 3,300 using the time-series model to 4,600 using the linear regression model (Figure ES1 and Table ES1).
A benefit of the Markov chain model is that it allows adjustments of the various parameters within the model to assess the impacts of various assumptions. Five scenarios demonstrated that there may be considerable variations in future demand for facility-based HD treatments depending on changes to policy and clinical practice. The Markov chain model can be utilised for future and ongoing assessment of the demand for renal dialysis.
Figure ES1: Demand projections for facility-based haemodialysis treatments, using three statistical models, Northern Territory 2013 to 2022
Table ES1: Demand projections for facility-based haemodialysis treatments, using three statistical models, Northern Territory 2013 to 2022
Year / Linear regression / Time-series / Markov chain2012 / 55,650 / 55,650 / 55,117a
2013 / 59,070 / 58,798 / 58,260
2014 / 63,048 / 62,062 / 60,967
2015 / 67,117 / 65,325 / 63,719
2016 / 71,315 / 68,588 / 66,154
2017 / 75,487 / 71,851 / 68,541
2018 / 79,812 / 75,115 / 70,969
2019 / 84,379 / 78,378 / 73,323
2020 / 89,360 / 81,641 / 75,704
2021 / 94,776 / 84,904 / 78,262
2022 / 100,367 / 88,168 / 82,185
aThe Markov chain model estimated the number of haemodialysis treatments in 2012 based on number of clients in 2012 but number of treatments in the previous years.
Renal Replacement Therapy Demand Study, Northern Territory, 2001 to 2022 25
Introduction
Across Australia, demand for renal replacement therapy (RRT) services has been growing at a significant rate over the last decade (1, 2). Health service funders for RRT are faced with increasing service delivery costs and investment requirements (3). The Northern Territory (NT) is experiencing growing demand for dialysis services. In 2009, the registered incidence of treated end-state kidney disease (ESKD) was 32 per 100,000 population (4). The number of patients on dialysis increased by 27% over the five year period from 2007 to 2011 (2). Same day haemodialysis (HD) now comprises close to 50% of total hospital admissions in NT public hospitals (5).
Although much could be done to slow the increase through targeted prevention and early intervention strategies, many health service funders have invested in greater capacity and new facilities only to find that actual demand significantly exceeded forward demand estimates and supply. New facilities that were thought to be sufficient for 20 years are now overflowing. The challenge for dialysis service planning is to manage priorities of equitable access with increasing demand for services in a financially constrained environment. Accurate projection of future demand is crucial for government in planning and providing adequate, effective and efficient RRT services for Territorians.
The objective of this study is to improve the understanding of projected demand and related service requirements to ensure future service sustainability through clinical service planning for dialysis services, where strategies could be developed to align available resources with demand.
This report includes:
· provision of a summary of ESKD for NT residents including the incidence of ESKD and treatments;
· projection of demands for RRT in the next 10 years including the application of different models and using different sources of data; and
· projection of demand for facility-based HD under several scenarios which model various assumptions that could arise from changes in policy and clinical practice.
Renal Replacement Therapy Demand Study, Northern Territory, 2001 to 2022 25
Methods
This study provides an estimate of NT RRT service provision from 2001 to 2012 and projects future service demand from 2013 to 2022. Baseline data was gathered about the numbers of patients with ESKD in the NT, including patients treated and not treated with RRT. Three different models were applied separately (time-series, linear regression and Markov chain) to project future facility-based dialysis demand.
Data sources
The study used four data sources:
· NT public hospital data: hospital admissions from 2001 to 2012;
· The Australia and New Zealand Dialysis and Transplant Registry (ANZDATA): NT data from 2003 to 2012;
· NT population data: NT population data 2003-2012 was used to estimate the ESKD incidence, prevalence and RRT modality; and
· NT population projection 2013-2020: The NT population projection data was sourced from Department of Treasury and Finance (6).
Statistical analysis
Descriptive analysis
The descriptive analysis provided in this study included the prevalence of ESKD and the proportion of those treated with RRT, compared to national averages. The ANZDATA registration rate was also estimated. The registration rate was used to adjust for the actual number of facility-based HD patients in projection models.
In this study, we first identified all ESKD patients from the NT public hospital data from 2001 to 2012. Two alternative definitions were used to identify patients with ESKD (Table 1). A “broad” definition followed the Australian Institute of Health and Welfare (AIHW) definition, which included both “specified”, and “unspecified” chronic renal failure (7). A “strict” definition excluded “unspecified” chronic renal failure, and focused on those with a clearly coded diagnosis of ESKD. Hospital data with all 49 diagnosis fields were used to identify cases with ESKD. Before the sixth edition of ICD-10-AM, introduced in 2008, earlier stages of chronic kidney disease (CKD) were not specifically recognised and as a result early or less severe CKD cases were coded as unspecified chronic renal failure. Both strict and broad definitions are presented in the report, as the former is more precise, while the latter allows comparison with national data published by AIHW.
We then estimated the proportion of ESKD patients receiving any type of RRT. Patients recorded with a diagnosis of ESKD were checked with procedure and diagnosis codes in the hospital data (Table 1). The date of the initial RRT and modality were identified. We compared the number of ESKD patients without recorded RRT to those with RRT in the NT, NT to the national rate, and Indigenous with non-Indigenous rates.
Palliative care provides conservative treatment for ESKD, rather than renal dialysis or transplant. We estimated the proportion of ESKD patients who died in palliative care in NT and compared this proportion with the national average.
Finally, survival curves for different RRT inception cohorts for both the NT and Australia were generated using the national ANZDATA data for the period from 1995 to 2011. The ESKD survival information was applied to subsequent modelling for the total RRT and satellite HD demand projections.
Table 1: Codes applied to identify end-stage kidney disease and treatment modality
ICD 10 diagnosis codes & description / ICD 10 procedure codes or block number / Other variablesESKD codes (broad definition) / N18.0 End-stage renal disease
N18.8 Other chronic renal failure
N18.9 Chronic kidney disease, unspecified
N18.90 Unspecified chronic renal failure
N18.91 Chronic renal impairment
N19 Unspecified renal failure
N18.5 Chronic kidney disease, stage 5
I120 Hypertensive renal dis w renal failure
I131 Hypertensive heart & renal dis w renal failure
I132 Hypertensive heart & renal dis w heart renal fail
ESKD codes
(strict definition) / N18.0 End-stage renal disease
N18.5 Chronic kidney disease, stage 5
I120 Hypertensive renal dis w renal failure
I131 H/T heart & renal dis w renal failure
I132 H/T heart & renal dis w heart renal fail
HD codes / Z49.1 Extracorporeal dialysis / 13100-00 / AR-DRG L61Z
PD codes / Z49.2 Other dialysis / Block: 1061, 1062
Kidney transplant / Z94.0 Kidney transplant status
Palliation / Z51.5 Palliative care / Care type
Notes: AR-DRG= Australian Refined Diagnosis Related Group; ESKD=end-stage kidney disease; HD=haemodialysis; ICD= International Classification of Diseases; PD=peritoneal dialysis
Linear regression model
In this study, the linear regression analysis was based on historical HD treatment utilisation rate, assuming that the rate increased with a linear relationship to the independent variables (e.g. age, gender and Indigenous status). Using yearly data from 2001 to 2012 from the NT public hospital separations dataset, the utilisation rate for the NT population was modelled for the subgroups broken down by Indigenous status, gender, and age group (under 40 years, 40 years and above).
For each group, the linear regression model was projected to 2022, taking into account projected NT population changes. In this study, we also reported 95% confidence interval to represent the upper and lower limits of statistical confidence.
Time-series model
The Auto Regression Integrated Moving Average (ARIMA) time-series model was also used to project demand for HD treatment. This method has been widely used in medical research, epidemiology and health economics, and has been shown to be useful in estimating future values when there is a significant random component. It is a form of regression analysis that seeks to project future movements by examining the differences and connections between the actual data and estimated values in the time-series. The advantage of the ARIMA model is that the related projections accommodate seasonal variation, errors and stationary periods within a dataset.
The model followed the same methods developed within the Department of Health in 2002 (8), which used monthly data on the number of HD treatments from NT public hospitals and expanded the model to project demand for facility-based HD treatments for Indigenous and non-Indigenous Territorians separately.
Markov chain model
A Markov chain model was also constructed as an alternative method for estimating RRT demand in the NT. The model was based on the general structure (including some assumptions) of an earlier model used to estimate RRT demand at a national level (3). The model follows a cohort of existing patients with ESKD and receiving RRT, along with the addition of patients starting RRT each year. For each cohort of future patients, the model assigned probabilities which were based on historical ANZDATA data from 2003 to 2012. The length of each ‘treatment’ cycle in the model was one year. The structure of the model is shown in detail in Figure 1, which presents the pathway for dialysis patients as they cycle through the different RRT modalities. The model was stratified by the age groups: 25-44, 45-64, 65-74, and 75 years and older. The Markov chain model used ANZDATA rather than NT public hospitals data.